September 30, 2013 at 7:25 am
L' Eomot Inversé (9/29/2013)
GilaMonster (9/29/2013)
Oh dear....If I may change the topic... For the purposes of a book chapter, what are your favourite statistics-related myths?
As 2013 is the International Year of Statistics, I'll assume you are talking about that science (or art) :alien: rather than about table statistics. :Whistling:
You might find the myths and realities[/o] page thatGuanajuayto U's maths research center put together for the international year of statistics helpful; or not, it may not be the sort of myth you are looking for. I have some more technical statistics myths, though:-
1) With a normal distribution there's a central band 3 standard deviations wide that contains 99.7% of the population. (It's actually 6 standard deviationw wide - the 3 sigma rule isays that 99.7% are within 3 standard deviations of the mean).
2) It's always safe to assume that everything has a normal distribution, because any other distribution is abnormal. That's just total nonense; plenty of things in the real world conform to Exponential distributions, Bounded Uniform distributions, Poisson distributions, Cauchy distributions, Weibull distributions, Pareto distributions, Erlang and other Gamma distributions, and hordes more.
3) The Shannon Hartley sampling theorem proves that you can only get 2B bits/second down a channel of bandwidth B Hertz, no matter how good the signal to noise ratio. (It actually says 2B real numbers to exact precision, ie 2B times an infinite number of bits; but I've had people with engineering degrees who worked in data communications making this claim, and been unable to convince them they were wrong; it says very little for the insititutions awarding those degrees that they delivered pig-headedly stupid graduates who didn't know the diffeence between a bit and a baud, and couldn't handle enough statistics to understand Shannon's encoding theorem).
4) You are always saffe using the law of large numbers. Well, you certainlty aren't when looking at specrum splitting and shifting due to electrostatic effects, or resonance energy distributions at the quantum level, or a few other bits of physics (some of it applied physics, not just theeory) where Cauchy distributions crop up.
5) Every statistical distribution has a mean and a variance; well, Cauchy distributions have neither, nor any higher moments, and that's why number 4 above is a myth.
6) If statistics predicts that something has low probability and then that something happens this proves statistics is wrong. (Interesting idea that; so low probability events don't happen? The favorite always comes first?) Theres also a similar one about high proability events that don't happen.
I think for this kind of statistics, the biggest myth of all is that you can't lie with them.
I dropped my microecon class in college after an argument with the TA where he insisted that not only couldn't you like with statistics, you couldn't make misleading graphs.
--------------------------------------
When you encounter a problem, if the solution isn't readily evident go back to the start and check your assumptions.
--------------------------------------
It’s unpleasantly like being drunk.
What’s so unpleasant about being drunk?
You ask a glass of water. -- Douglas Adams
September 30, 2013 at 7:26 am
Jack Corbett (9/30/2013)
Jeff Moden (9/28/2013)
GilaMonster (9/28/2013)
Jeff Moden (9/27/2013)
Speaking of "Are the posted questions getting worse"?... I was going through some old responses to some posts and I came across this one that no one responded to...After googling, i have learnt to open a new clr stored procedure project to convert rtf format. But as i told, i have no knowledge..So could you please guide me from here?
First, I congratulate everyone for not responding to that post because even I have my limits and had to make a conscious effort to not respond to it with less than subtle hints about using tools they don't know anything about and maybe a rant about having other people do their job for them.
The economy must be picking up because I'm seeing more and more "I have no knowledge {of SQL}" types of posts across the board. Apparently, more and more people are getting themselves hired with "no knowledge". I blame the interviewers for this type of thing.
Have you folks observed the same kind of thing over the last month or so or is it just me?
Have you seen the myriad of dynamic SQL related questions the last few days?
Heh... between that and TempDB questions (and some of the answers, apparently... 50MB) by people who have "no knowledge", it's been a real hoot.
I would say that it depends on where you are. In New England the unemployment rate is ~2% in IT and there are a lot of SQL Server jobs open. I'd also venture to say that the 2% are probably not top-notch or are new graduates who have unrealistic compensation expectations for their first job. Employers are still trying to low-ball people, which is why positions stay open in New England. No ones moving to a new job without a decent increase.
Also in the US, pretty much all the consultants I know have enough work and are actually looking for help in many cases. Case in point is Brent Ozar Unlimited adding an employee.
I should probably touch bases with you toward the later half of November early part of December if I am not seeing anything back home. I already told Betty I would move if I needed to in order to find work for me. May make reconciliation a little more difficult, but we are willing to make it work if needed.
September 30, 2013 at 8:15 am
Anyone want to help on the blocking thread?
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
September 30, 2013 at 8:48 am
GilaMonster (9/30/2013)
Anyone want to help on the blocking thread?
I would take a look at it if I had a link to it.
September 30, 2013 at 8:48 am
Lynn Pettis (9/29/2013)
I am officially a grandfather, again. Kassondra gave birth to her second son, Branson James on 9/28 @ 6:19 PM MDT!
Congrats!
Follow me on Twitter: http://www.twitter.com/way0utwest
Forum Etiquette: How to post data/code on a forum to get the best help
My Blog: www.voiceofthedba.com
September 30, 2013 at 8:57 am
Jack Corbett (9/30/2013)
GilaMonster (9/27/2013)
Thursday at Summit is going to be sooo funny. Just saying...Okay, I need some more info on this one. I don't see anything on the schedule that funny going on Thursday.
Uhh . . . Kilt Thursday, maybe?
Jason Wolfkill
September 30, 2013 at 8:58 am
wolfkillj (9/30/2013)
Jack Corbett (9/30/2013)
GilaMonster (9/27/2013)
Thursday at Summit is going to be sooo funny. Just saying...Okay, I need some more info on this one. I don't see anything on the schedule that funny going on Thursday.
Uhh . . . Kilt Thursday, maybe?
Or perhaps the good-natured mocking of those who dare to take the mic at #sqlkaraoke the night before?
Jason Wolfkill
September 30, 2013 at 9:48 am
Brandie Tarvin (9/30/2013)
GilaMonster (9/30/2013)
Anyone want to help on the blocking thread?I would take a look at it if I had a link to it.
The one in recent posts with the stuck shift key.
http://qa.sqlservercentral.com/Forums/Topic1499887-146-1.aspx
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
September 30, 2013 at 11:14 am
Accidental duplicate post deleted.
Tom
September 30, 2013 at 11:15 am
Stefan Krzywicki (9/30/2013)
I think for this kind of statistics, the biggest myth of all is that you can't lie with them.I dropped my microecon class in college after an argument with the TA where he insisted that not only couldn't you like with statistics, you couldn't make misleading graphs.
Can you lie with graphs? Yes. In fact they might almost have been invented to assist people to lie. Can you get away with it? Only if your audience is not bright enough to spot your tricks. Starting the Y axis at 1000000 instead of at 0 so as to make a variation of 3 units look like a 300% change instead of a 0.0000003% change is so obvious; and leaving the Y axis without any scale on it is very obviously a signal that you've done it and are trying to conceal it; of course there are less obvious tricks too, but a little bit of critical common sense will enable people to spot them. I'm not sure what proportion of audiences is going to be fooled, but I suspect it's not terribly big for most of the common tricks with graphs.
Can you lie with statistics? Yes. Can you get away with it? Only if your audience is not educated in statistical thinking or not bright enough to catch you out. The trouble here is that so many people think they understand statistics when they haven't a clue about statistical thinking or probabilities or uncertainty, so that sometimes you can get away with it.
Is there anything you can't lie with? No. So what's so special about graphs or statistics that you have to ask the question about them?
That teaching assistant probably needed asistance in understanding statistical thinking; maybe even with finding the usual part of his anatomy, even when trying with both hands; more importantly being told to stop peddling pure nonsense or quit being a TA.
Tom
September 30, 2013 at 11:36 am
L' Eomot Inversé (9/30/2013)
Stefan Krzywicki (9/30/2013)
I think for this kind of statistics, the biggest myth of all is that you can't lie with them.I dropped my microecon class in college after an argument with the TA where he insisted that not only couldn't you like with statistics, you couldn't make misleading graphs.
Can you lie with graphs? Yes. In fact they might almost have been invented to assist people to lie. Can you get away with it? Only if your audience is not bright enough to spot your tricks. Starting the Y axis at 1000000 instead of at 0 so as to make a variation of 3 units look like a 300% change instead of a 0.0000003% change is so obvious; and leaving the Y axis without any scale on it is very obviously a signal that you've done it and are trying to conceal it; of course there are less obvious tricks too, but a little bit of critical common sense will enable people to spot them. I'm not sure what proportion of audiences is going to be fooled, but I suspect it's not terribly big for most of the common tricks with graphs.
Can you lie with statistics? Yes. Can you get away with it? Only if your audience is not educated in statistical thinking or not bright enough to catch you out. The trouble here is that so many people think they understand statistics when they haven't a clue about statistical thinking or probabilities or uncertainty, so that sometimes you can get away with it.
Is there anything you can't lie with? No. So what's so special about graphs or statistics that you have to ask the question about them?
That teaching assistant probably needed asistance in understanding statistical thinking; maybe even with finding the usual part of his anatomy, even when trying with both hands; more importantly being told to stop peddling pure nonsense or quit being a TA.
What was so special about them at the time was how he was showing the class how the data led directly to this graph and only this graph and how it clearly displayed the data without ambiguity.
--------------------------------------
When you encounter a problem, if the solution isn't readily evident go back to the start and check your assumptions.
--------------------------------------
It’s unpleasantly like being drunk.
What’s so unpleasant about being drunk?
You ask a glass of water. -- Douglas Adams
October 1, 2013 at 1:26 am
Just caught up on The Thread again after a while... and yes, there's a new lot who can't even press f1 or Google for help on something as basic as converting strings to dates.
Surely it's quicker to look it up yourself than to post a question on a forum and wait for an answer which probably tells them to look it up anyway??? :unsure:
And congrats to Lynn. I can understand his staying on in Afghanistan as long as possible, as folks here seem to be enjoying the life and they're putting in for second tours once they are back despite an enforced year break between tours.
There seems to be work around in the UK for SQL DBA/developers - I keep getting good posts in agency emails despite not having registered with them for at least seven years!
October 1, 2013 at 3:27 am
Could somebody with more experience with Execution Plans have a look at this thread please - http://qa.sqlservercentral.com/Forums/Topic1498743-23-1.aspx ?
Because I've never used OPENQUERY outside of idle curiosity I don't know if the Remote Scan is a feature of it or if it is the Implicit Conversions that abound within the plan.
And could anyone explain to me why there's an Implicit Conversion for nVarchar(MAX) when both fields are nVarchar(MAX)? I've never noticed that behaviour before.
Thanks.
October 1, 2013 at 6:11 am
Well, should be a rather interesting next couple days / weeks / (REALLY hope not) months with the US Government going into shutdown.
As some of you know, I'm now a contractor at a govmn't facility. Really weird watching so many of the government employees heading home already. I thought things were bad when people were furloughed for one day a week because of the sequester a couple months back, now it's going to be an all week long furlough...
October 1, 2013 at 6:18 am
jasona.work (10/1/2013)
Well, should be a rather interesting next couple days / weeks / (REALLY hope not) months with the US Government going into shutdown.As some of you know, I'm now a contractor at a govmn't facility. Really weird watching so many of the government employees heading home already. I thought things were bad when people were furloughed for one day a week because of the sequester a couple months back, now it's going to be an all week long furlough...
The USA government are a bunch of crybabies. 🙂
At the latest election in Belgium, we didn't have a government for 541 days, and nobody even noticed. Everything kept functioning like normal...
Need an answer? No, you need a question
My blog at https://sqlkover.com.
MCSE Business Intelligence - Microsoft Data Platform MVP
Viewing 15 posts - 41,521 through 41,535 (of 66,000 total)
You must be logged in to reply to this topic. Login to reply